Deep Learning-Based Anomaly Detection to Classify Inaccurate Data and Damaged Condition of a Cable-Stayed Bridge

Abstract

Cables of cable-stayed bridges are gradually damaged by weather conditions, vehicle loads, and corrosion of materials. Stayed cables are an essential factor closely related to the stability of a cable-stayed bridge. Damaged cables might lead to the bridge collapse due to tension capacity lost. Therefore, it is necessary to develop structural health monitoring (SHM) techniques that check the cable conditions. Besides, the sensor network system development has contributed to the state analysis, such as damage detection and structural deformation, by allowing us to collect large-scale SHM data. However, the collected SHM data might include abnormal data due to device malfunctioning or unexpected environmental inconstancies. Furthermore, since data anomalies interfere with accurate structural evaluation, we need to identify anomalies and treat them appropriately in the data preprocessing stage. However, the cause of anomalies may be either temporary errors or actual structural deformation. These anomalies are informative data that enable us to discover damages to the structures. In this paper, we distinguish between anomalies as inaccurate data and anomalies related to the state of structures and propose a framework to identify each of them. We train a Long Short Term Memory (LSTM) network based Encoder-Decoder architecture that processes multivariate time series and learn temporal correlation. The trained LSTM network discovers anomalies by calculating anomaly scores. We determine the anomalies emerging intermittently as errors and correct the erroneous data. If the anomalies persist, we recognize the data as generated by bridge damage or sensor device failure. We evaluate the proposed technique with cable tension data from an actual cablestayed bridge.

Unsupervised Learning Methods for Data-Driven Vibration-Based Structural Health Monitoring: A Review

“…A long short-term memory (LSTM) network is a type of recurrent neural network that is commonly used to learn complex temporal patterns from time-varying data [ 174 , 175 ]. Son et al [ 84 ] proposed a two-stage anomaly detection framework, relying on an encoder-decoder LSTM network, for identifying abnormalities in the collected SHM data. Since their focus was on monitoring cable-stayed bridges, the input to their two-layered LSTM network is the raw cable tension time series, and the reconstruction error is used for estimating an anomaly score.…”

Deep Learning for Structural Health Monitoring: Data, Algorithms, Applications, Challenges, and Trends

“…Among them, data cleaning is the process of anomaly detection and removal, which is the focus of the current research. Son et al [67] developed an LSTM-based encoder-decoder architecture to process time series data and calculate anomaly scores. Then, temporary errors were identified and removed according to the abnormal scores.…”

“…In addition to CNN, Recurrent Neural Networks (RNN) can identify the time features of data. Through Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), it can not only realize multi-type damage identification and location [61], crack detection [62,63], and dam displacement monitoring [64][65][66] but also detect data anomalies caused by sensor faults or environmental changes [67][68][69], and even predict missing data [70][71][72][73]. There are also unsupervised learning algorithms, such as auto-encoder and Generative Adversarial Networks (GAN), which directly use unlabeled data for feature extraction.…”

Artificial Intelligence and Structural Health Monitoring of Bridges: A Review of the State-of-the-Art

“…A traditional CNN structure for classification is shown in Fig 5 [155]. The high capability and efficiency of deep learning networks in adapting to various issues and complexities in SHM of bridges, as well as their ability to function properly in learning from large amounts of data, has led to valuable studies in this field in recent years [156][157][158][159][160][161][162][163][164]. For bridge damage detection, Fernandez-Navamuel et al, (2022) developed a supervised deep learning strategy that incorporates Finite Element models to enhance the training phase of a deep neural network.…”